Step 01
Ingest
The worker container syncs markdown and PDF from Google Drive, parses content, and generates 384-dim embeddings with MiniLM-L6-v2.
MySoul · v0.1
Conversational capability profiles.
MySoul is a local-first, multi-tenant platform that turns a person's documents into an evidence-backed profile visitors can interview. Ask a question, get a grounded answer with citations — or browse the tenant directory to find who you're looking for.
Start here
Choose a profile, inspect the evidence, then ask better questions.
How it works
From documents to conversational profile
Each tenant's Google Drive folder is synced, parsed, and embedded. When a visitor asks a question the retrieval pipeline finds the most relevant chunks and synthesizes a grounded answer.
Step 02
Retrieve
When a visitor asks a question, the chat endpoint embeds the query and runs pgvector cosine similarity search scoped to the tenant.
Step 03
Answer
The system synthesizes an evidence-grounded answer with inline citations — extractive template mode by default, or LLM mode via endpoint config.
Chat request flow
POST /api/chat/[slug] → resolve tenant by slug → embedQuery (MiniLM-L6-v2, 384-dim) → searchContent (pgvector cosine, tenant-scoped, top-5) → synthesizeAnswer (template extractive with citations, or LLM mode via LLM_ENDPOINT) → return { answer, sources }.
Platform
Multi-tenant capability profiles
Visitor flow
Browse → profile → ask questions
Retrieval
pgvector cosine similarity
Evidence
Grounded answers with citations